Industry-level Expenditure on Intangible Assets in the UK*

Transcription

1 Industry-level Expenditure on Intangible Assets in the UK* Valentina Gil Queen Mary, University of London and CeRiBA Jonathan Haskel Imperial College Business School; CeRiBA, CEPR and IZA JEL reference: O47, E22, E01 Keywords: intangible assets, R&D, training, organisational capital, investment October 2007 Revised, November 2008 Abstract We present data on expenditure on intangible assets for the UK market sector for six industries over the last decade or so (data availability allowing). The sectors are (1) Agriculture, Fishing and Mining; (2) Manufacturing; (3) Electricity, Gas and Water; (4) Construction; (5) Wholesale and Retail, Hotels and Restaurants, Transport and Communications; (6) Financial Intermediation and Business Services. We use new methods relative to previous work particularly in advertising and design. Our main findings are as follows. First, overall intangible expenditure was, in 2004, around 152bn (investment was 92bn). Second, to give an illustration of the manufacturing results, in 2004 manufacturing intangible investment was 30bn and tangible 12bn and intangible, a ratio of 2.5 to 1 up from 1.2 to 1 in So whilst the ratio of manufacturing tangible investment to value added has been falling, the ratio of manufacturing intangible investment to value added has been rising, leaving the overall ratio about the same. Third, manufacturing accounted in 2004 for 12% of total tangible investment and 31% of total intangible investment. *Contact: Jonathan Haskel, Imperial College Business School, Imperial College, London SW7 2AZ,:, j.haskel@ic.ac.uk. Financial support for this research comes from the Department for Business, Enterprise and Regulatory Reform (BERR), ONS and the COVEST project, < funded by the European Commission Seventh Framework Programme, Theme 9, Socio-economic Science and Humanities, grant number , we are grateful to all institutions concerned for their support. This work was carried out at CeRiBA and ONS, the Centre for Research into Business Activity at the ONS. This work contains statistical data from ONS which is Crown copyright and reproduced with the permission of the controller of HMSO and Queen's Printer for Scotland. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which may not exactly reproduce National Statistics aggregates. We are very grateful for comments to our fellow team members at the ONS of Tony Clayton and Fernando Galindo-Rueda and of colleagues on the Project Steering Group: Angus Armstrong, Stuart Barthropp, Ivan Bishop, Paul Crawford, Ian MacCafferty, Richard Stead, Gavin Wallis, Ken Warwick and David Wilkinson. We are also grateful to Mauro Giorgio Marrano, Graeme Chamberlin, Sumit Dey-Chowdhury, Peter Goodridge and Joe Robjohns for help with data. All opinions and errors in this paper are our own. 1

2 1. TRODUCTION This paper reports estimates of investments in intangibles by industry following the whole economy work in the US by Corrado, Hulten and Sichel (2002, 2004, henceforth CHS) and the UK by Giorgio Marrano, Haskel and Wallis (2007, GHW). We use mostly the same sources of data and methods followed by CHS and GHW and additionally we attempt to estimate expenditure on intangible assets at an industry level instead of at the all private sector level. Table 1 summarises the sources and the industry split available. The first column shows the categories of intangible assets used by CHS. The second column reports the sources which are a mix of National Accounts, of official surveys and estimates from other sources. The third column shows the industry breakdown and the fourth column reports time series data availability. The rest of this paper is as follows. The next section explains how we get a common industry split for all the intangibles, the third section gives details of the specific procedure followed for the estimation of expenditure for each intangible asset. The fourth section presents further analysis for the manufacturing sector. The fifth section concludes. The appendix discusses in more detail some conceptual points such as our foreign trade adjustment, treatment of licences etc. Finally, all data from this paper are available for download from the COVEST website ( THE DUSTRY BREAKDOWN To measure expenditure on intangibles at an industry level we return to the original data sources used to compile the whole market sector data. Some of these sources provide an industry breakdown, but industry split they use does not always relate to the official industry classification. This is because some of the surveys are not ONS officially collected data, or have been collected for purposes requiring a non-sic division, which means they do not necessarily use the SIC classification. Based on 1 We are very grateful to Annarosa Pesole for preparing the data and programmes in accessible form. 2

3 the existing sources, we propose a structure that combines the available sources at the lowest common possible level of aggregation. Table 2 illustrates how the sectoral classifications for key intangible investment items relate with each other. Column 1 shows the industries corresponding to the Input/Output definitions, which is the source of the data for software (purchased and own-account). Column 2 shows the industry breakdown for the training data, from the NESS05 (see below). These data are collected for the Learning and Skills Councils and are therefore provided at that level of industry disaggregation, which we have mapped into the SIC. Column 3 shows the industry breakdown of the R&D data, from the BERD Inquiry and column 4 that of the purchased organisational structure data. The latter is spending on knowledge from the Management Consultants Association and is quite aggregated. By pooling these sources, a common classification involving six main sectors emerges. These are depicted using different shades and summarised in Table 3. As table 3 shows, we have six feasible categories: 1) Agriculture, Fishing and Mining 2) Manufacturing 3) Electricity, Gas and Water 4) Construction 5) Wholesale and Retail, Hotels and Restaurants, Transport and Communications 6) Financial Intermediation and Business Services We exclude sections L (Public Administration), M (Education), N (Health), O (Personal services), P (Private Households) and Q (Extra-Territorial). These involve activities mostly outside the market sector for which we have no reliable output or other input data. Regarding further disaggregation of manufacturing, first, a more detailed disaggregation for the manufacturing sector does exists for the categories other than organizational investment at around a two-digit level. Nevertheless, it is possible to provide a more detailed disaggregation of intangible investment in the manufacturing sector for software, R&D, training, advertising, market research, own-account expenditure on organizational structure and design. 3

4 2. ESTIMATION OF VESTMENT BY TANGIBLE ASSET CLASS 2.1 COMPUTERISED FORMATION Computerised information comprises computer software, which includes purchased and own account software development, and the value of new computerised databases. The main source for computer software investment is the work already carried out by the ONS, described in Chamberlain, Chesson, Clayton and Farooqui (CCCF, 2006 and CCCF, 2007). Estimates of purchased software are based on data from three different company investment surveys: Business Spending on Capitalised Items (BSCI), the Capital expenditure Survey (Capex) and the Annual Business Inquiry (ABI). For own account spending, estimates are based on the earnings of employees in computer software occupations, using the Annual Survey of Hours and Earnings (ASHE). CCCF chose the occupations of ICT managers, IT strategy and planning professionals, software professionals, IT operations technicians, user-support technicians, database assistants/ clerks and computer engineers and installation and maintenance personnel. They estimated headcounts and wages, upwards adjusted the numbers to reflect the full employment cost and then downward adjusted them to reflect the fractions of time spent on development versus maintenance. A final adjustment is made to reflect possible sales to other firms that would imply double counting. To avoid any double counting we do not consider any additional spending on computerised databases as the ONS software figures already factor these in - two of the three computer purchase surveys asked firms to include database spending as part 4

5 of software spending and the own account data include the wages of database assistants and clerks. Regarding the industry breakdown of software expenditure, both own-account and purchased software time series are available at the 123 industry level. Figure 2.1 shows the total expenditure on software from 1970 to The private sector expenditure has gone up in recent years while the manufacturing expenditure growth accounts for only a small proportion of the total increase. In 2005 the manufacturing sector spent 2.6 bn on software, while the overall private sector spent bn. The financial and business sector is responsible for almost the half of this total. Fig Software expenditure by sector Tot FinBsSvc ReHtTrn Mfr year Tot Mfr Cons FinBsSvc AgMin Util RetHtTrn In May 2007 the ONS has revised down the estimates on software expenditure (see CCF, 2007). The previous private sector total expenditure for 2004 showed in Giorgio Marrano, Haskel, Wallis (GHW) (June 2007) was 21.59bn, the new total is 16.5bn. 5

6 In the National Accounts the total software investment is estimated to be about 19.5 billion in However this includes about 1.2 bn of public sector purchased investment and also a small amount of public sector own-account. Taking away these portions the private sector software investment is bn. The difference between this total and our total presented above ( 16.5bn) is due to the different method we used to estimate the private sector. As explained in page 3 we compute the private sector excluding sections from L to Q, therefore we are not including the private sector investment on software of these industries. Table 4 shows the data for 2004 only, with the expenditure data for each sector and the fractions of overall software spending accounted for by sector. Spending by Financial and Business services dominates this intangible investment class at 48% of total spending, with Retail etc. accounting for 31% NOVATIVE PROPERTY Scientific Research & Development Expenditure data on R&D performed by businesses in the UK are derived from the Business Enterprise R&D survey (BERD). This survey complies with the OECD international standards set out in the Frascati Manual (FM). Although the FM definition of R&D is more general, BERD questionnaires are intended to capture R&D aiming to resolve scientific and technological problems. Items such as design, market research are explicitly excluded. Double counting of R&D and software investment is a potential problem. Firms in the computer industry asked to report their R&D are advised as follows: For software development to be classified as R&D, its aim must include the resolution of scientific or technological uncertainty on a systematic basis. Routine software is not R&D. The use of software for a new application or purpose does not by itself constitute R&D; the application must be significantly different and resolves 6

7 uncertainties of general relevance. Software development within an R&D project should be classified to the product sold by your company that makes use of the software in its manufacture or within the product itself. For example work on software to be used within a motor vehicle engine would be allocated to the motor vehicle product group. Software which is developed and sold as software for direct use by customers should be allocated to product group AE (computer and services). We therefore decided to subtract R&D spending in the computer and related activities industry (SIC 72) from the R&D spending figure of the financial intermediation and business service sector (our sector 6) to avoid potential doublecounting. The BERD data exclude R&D purchased from companies abroad and include R&D undertaken for companies abroad. In order to derive a measure of UK investment we subtract R&D exports and add R&D imports, see section 4 for more details on this. In brief, first, we collect the R&D exports and imports time series from the ONS. Fig a illustrates the exports and imports time series. Fig a bn 5 R&D International Trade R&D Exports R&D Imports Second, to allocate exports and imports by sector we compute the sector-specific imports share: the Supply Table 2.1 in the Blue Book shows the imports of goods 7

8 and services by sector. We compute the ratios between the total imports of all goods and services for each sector and the total imports for all goods and services by the private sector. We then apply these ratios to both the exports and imports time series by assuming that the sector-specific export-shares are equal to the sector-specific import shares 2. Finally, we subtracted exports and added imports to the BERD R&D industry data (for more on the reason for this, see the section below). The industry split of R&D is provided by ONS. However, a number of points are worth making regarding the accuracy of the industry split. First, the provided industry split relates to the main product produced by enterprises rather than the activity sector of their establishments (activities are preferred in National Accounts guidance). Sector SIC73 is specifically designed to cover the delivery of R&D services and SNA93 suggests that vertically integrated companies who do R&D and final production should provide separate data for these separate activities, the former of which should be allocated to SIC73. For example, a pharmaceutical company that undertakes substantial R&D would be asked for National Accounting purposes to allocate its R&D activity to a separate establishment belonging to the R&D sector. Its output will be intermediate consumption of the main establishment in the Pharmaceutical sector. BERD will record this company s R&D as corresponding to the Pharmaceutical sector because that is the main product. This example poses no misattribution problems because it is the pharmaceutical sector that genuinely undertakes the investment. Second, because of the way that BERD is collected the industry doing the R&D is the industry that is performing the R&D. Every firm who claims to be undertaking R&D expenditure (in a simple yes/no question) on the ABI is then surveyed on the BERD. So a vertically integrated aerospace company who spends 100 of in-house R&D means the manufacturing does 100 of R&D and services none. If it then spins off the R&D to an independent company, then manufacturing R&D falls to zero and services rises to 100. An alternative method is to try to calculate the R&D undertaken by sector of funding and not performing. So if the original company 2 We follow the same procedure in doing the trade adjustment for purchased organizational structure. 8

9 funded the now-independent company for 100, then in terms of the sector funding the R&D the sectoral allocation would not change. Figure 2.2.1b shows the evolution of the six sectors expenditure on R&D. This confirms that manufacturing is the main contributor to overall scientific R&D. Figure 2.2.1b Scientific R&D expenditure by sector Mfr Tot year Tot Mfr Cons FinBsSvc AgMin Util RetHtTrn As it is illustrated in the graph, the total expenditure on R&D for 2004 is 9.11bn whereas in GHW this figure is 12.4bn. However, this difference is mainly due to our adjustment for the international trade. In fact, before doing this adjustment the total expenditure for 2004 is 12.27bn. When we try to measure the UK investment in R&D we end up with a lower value for total spending as a significant amount of the performed R&D captured by the BERD is actually R&D undertaken for foreign companies (R&D exports). Table 4 sets out the shares of R&D by sector, including the trade adjustment, for 2004 and confirms that manufacturing does the lion s share of measured scientific R&D. Its share is 83%, which differs from the figure of about 75% which is often 9

10 quoted, derived from R&D total expenditure in 2005 at about 13bn and manufacturing being about 10bn, which is 76%. But we subtract spending on computer and related activities so as not to double count software. Since software is in the service sector, then the manufacturing share is about 10/12=83% Mineral Exploration Expenditure on mineral exploration is already counted as an investment in the National Accounts. It includes the cost of drilling and related activities such as surveys. As it is the R&D undertaken by the mining sector only, we do not provide an industry split. To estimate this expenditure we add the time series of coal mineral exploration, continental shelf companies and mineral exploration other than coal and oil. These estimates are the same as presented in GHW. Fig shows the trend of the expenditure on mineral exploration. Fig Mineral exploration expenditure bn Further details of this sector are as follows, see National Accounts, Concepts, Sources and Methods (NACSM) (2006, para 15.60ff). For mineral oil and natural gas, estimates of GFCF, including exploration costs, are based on information collected in the DTI quarterly inquiry into the oil and natural gas industry. The estimates cover fixed capital formation by exploration licensees, production licensees, operators appointed by production licensees and specialised contractors selling services to the industry. The fixed capital formation of contractors who may operate in the North Sea or other oil fields is included only where the contractors are 10

11 registered UK companies. Where operators work as part of a consortium, expenditures are allocated in proportion to their shares. For other mining and quarrying, estimates of GFCF, including exploration costs, in the coal industry are derived from information collected in ONS quarterly capital expenditure inquiries Copyright & license costs The National Accounts records this intangible item as an investment. The Source and Methods Manual says The production of books, recordings, films, software, tapes, disks, etc, is a two-stage process of which the first stage is the production of the original and the second stage the production and use of copies of the original. The output of the first stage is the original itself over which legal or de facto ownership can be established by copyright, patent or secrecy. This is recorded as capital formation. The value of the original depends on the actual or expected receipts from the sale or use of copies at the second stage, which have to cover the costs of the original as well as costs incurred at the second stage. We estimate copyright and license costs by adding up the time series of: artistic originals: broadcasting and recording, Entertainment, literacy and artistic originals: Public Corporations, Artistic originals: Publishing. The last one can be accounted in the manufacturing sector whereas the first two are in Recreational, Cultural and Sporting Activities (SIC 92). These are out of the market sector according to our definitions. These estimates are the same as presented in GHW. Fig shows the manufacturing expenditure compared to the total expenditure in copyright and license. Fig

12 Copyright & License Costs expenditure bn Manufacturing Total Private Sector Further details of this sector are as follows, see NACSM, para 15.60ff. They state that to measure the investment due to the production of originals in this sector, the data relate only to the TV and radio, publishing and music industries. The BBC is included. NACSM notes that estimates are based on information from BBC annual reports and various other sources. No detailed description is provided on how the relation between spending on originals and copyrights, licenses etc. translates into investment. OTHER PRODUCT DEVELOPMENT, DESIGN AND RESEARCH New product development costs in the financial industry Following CHS we measure new product development in financial services as 20% of total intermediate consumption by the financial services industry, taking the data from the Use Table. This intermediate spending includes also the purchase of advertising, software, consulting services and architectural and engineering activities which are counted elsewhere in the spending calculations. Therefore we subtract the purchase amount and take the 20% of this adjusted figure. These estimates are the same as presented in GHW. Fig shows the results of our estimation. 12

13 Fig bn New Product Development costs in the financial industry New architectural and engineering designs In Galindo-Rueda, Haskel and Pesole (GHP from now on, 2008) we set out results for own-account and expenditure in Architectural and Engineering Design (AED) for the UK market sector. We use these numbers in that paper here. 3 Briefly, in GHP we exploit the information from the SU tables to obtaining improved figures for both own-account output and the purchased output attributable to the AED industry. We do that using information from several sources (i.e. ABI, Business Structure Database, ASHE) to scale down the total output reported by the SU tables for industry 112, which represents a broader category than the AED industry. As in this paper, in GHP we use the turnover share from ABI to remove SIC 74.3 Technical testing and analysis from industry 112. In addition, we remove some subsector of SIC 74.2 Architectural and engineering activities and related technical consultancies in which AED activities do not incur. 3 The expenditure numbers in GHP are larger than those originally reported in the earlier version of this GH paper and indeed in the GHW paper, but the investment numbers are very similar. The reason for the difference is the following. In GHP we use the software method of estimating own-account spending by industry via labour market data on design occupations by industry. We did not do this in GHW, but, following CHS, halved all purchased design to estimate total purchased and own-account. In the earlier version of this paper we also used labour market data, but used a broader set of occupations than we do in GHP and assumed they were more costly. Finally, as we explain in GHP although we have higher spending data than GH, our design survey suggests that we should allocate a lower level to investment, leaving us with about the same investment levels as in GHW and the earlier GH paper. 13

14 First, we estimate the purchased component from the IO Tables. We derive the AED purchased output as the sum of Intermediate Consumption (IC) and Gross Fixed Capital Formation (GFCF). We reallocate IC and GFCF between products according to the import and export data reported in the Pink Book (2005). We use these data to fully balance the supply and use side. This method is consistent with the construction of the SU tables. Fig a illustrates the expenditure on purchased AED by industry, showing financial and business services at about half of this group. 14

15 Fig a Purchased AED output by industry, AED purchased output, m Mfr Cons Util Total FinBsSvc RtHtTrn AgMin year AgMin Util RtHtTrn Total Mfr Cons FinBsSvc To measure the AED own account, looking at the SU tables, we construct a hypothetical cost-structure that will incur in the production of in-house AED. Implementing these adjustments to our initial figure for the output of industry 112 we obtain the output for the AED industry, which we denote Y 22. Thus, we can now determine the AED output outside the AED industry in industry i as: Y Qi = WN WN 22 occ = AED occ = AED ind= i ind = AED where both the wage bill for designers inside and outside the AED industry are revised for an estimate of the time spent in AED activities and the number of self employed. A last adjustment is needed to obtain the AED own account. We need to subtract to the AED output reported above the amount of AED output sold in the market place. 15

16 Doing so we avoid any double counting since the latter is already accounted by National Statistics. In 2004, we find that own account spending in AED is about 27m. Private sector spending on purchased AED is about 17m. Thus, our final figure for the total expenditure in AED is roughly 45m. Note that these are expenditure data: we shall multiply these by 50% to obtain investment, relying on the survey evidence in GHP. Finally, in Figure b the left hand figure shows the own account spending by industry and the right hand figure the spending for purchased and own account in the whole economy. A number of points regarding the figure are worth noting. First, the data for 2005 and 2006 are provisional, being extrapolations of the 2004 data since the 2005 and 2006 IO tables are not available. Second, manufacturing accounts for the lion s share of total spending, although the trend in Finance and Business Services is upward. 16

17 Fig b Expenditure on AED by industry, design own account, m RtHtTrn Mfr FinBsSvc Util Cons year AgMin Util RtHtTrn Mfr Cons FinBsSvc design output, m total oa purchased year total purchased own-account R&D in social science and humanities As in GHW it is estimated as twice the turnover of R&D in the SIC 73.2 Social Sciences and Humanities, with the doubling being assumed to capture own-account spending. Fig shows the trend of the expenditure on this intangible item. 17

18 Fig bn R&D in Social Sciences & Humanities ECONOMIC COMPETENCIES EXPENDITURE ON BRAND EQUITY Advertising We estimate advertising expenditure from the IO Tables. The Use table 3 provides intermediate consumption in Advertising (product group 113) by industry (at the 123 industry level). Thus we take the sum of these purchases across all industries. In this case there is no value to add from the Gross Fixed Capital Formation (GFCF). A number of points are worth noting. On one hand it is likely that the figures obtained underestimate the investment in advertising as we are not capturing the own-account component. On the other hand they include the classified advertising (i.e. small advertising appearing at the end of newspapers typically for small items of sale or vacancies) which is unlikely to be asset building. Figure shows the expenditure trend on advertising undertaken by all sectors. It is evident that advertising expenditure for both private sector and Financial & Business services has grown in the recent years and is valued respectively at 17.5 billions and at 7.54 billions in

19 Figure Advertising expenditure by sector Tot FinBsSvc RetHtTrn Mfr year Tot Mfr Cons FinBsSvc AgMin Util RetHtTrn Comparing these results with GHW we do not find similar estimates: in GWW the total expenditure on advertising for 2004 is 14bn, in the current work is 17.5bn. Financial and Business Services sector accounts for almost 50% of this expenditure. Note in passing that a previous version of this paper found a somewhat smaller total ( 12bn, but this excluded classified advertising), but a quite different distribution of advertising expenditure across industries. That version used Advertising Association data on spending on billboards, TV and internet etc. that was disaggregated by industry. On those data, manufacturing industry spent about 52% of total advertising. Part of the difference might be due to the implicit inclusion of ownaccount advertising in the Advertising Association data, but part might be due to differences in classification in that data. We must however stress caution on the attribution of advertising to industries. First, an advertisement for cheap product X in supermarket Y might be building a brand for the product or the supermarket. Second, there is a generic problem of assigning any activity, be it advertising or anything else, to a firm who does both service and manufacturing activities. Suppose that an integrated firm A makes cars and 19

20 employers cleaners. To what industry is that firm assigned? ONS have completed a recent paper on this issue and it turns out to be not straightforward. The general treatment of a firm involved in different activities is to split the firm up into different reporting units and ask them to report on activities separately, so that firm A will have one aggregated firm record but two reporting unit records. Each reporting unit is assigned to the industry in which it belongs, but the firm as a whole is assigned into the industry which the majority of employment is located in, which in the firm A example is manufacturing. This then gives the well-known problem that if there is contracting out of services, manufacturing apparently shrinks. If the manufacturing part of the firm keeps on advertising, then the allocation of advertising to this firm does not change, but the share of employment that advertising is compared to does. A further slightly different example is as follows. Suppose integrated firm B has a distribution arm and a manufacturing arm, but with employment such that the overall firm is manufacturing and advertising is classified there. Next suppose that it disintegrates (and suppose sends the manufacturing abroad). Then the advertising is assigned to services. The way to make this consistent might be to have the original breakdown by reporting unit, in which case the employment share of manufacturing, in this example, was previously overstated. Note that this would have to be done for the service sector too: consider for example a retail bread shop with a bakery in the store. This is classified as services on a firm basis, if most of the staff work on retailing, but on a reporting unit basis should have been part manufacturing. The likelihood is that the reporting unit records are not robust enough to track this and so a consistent assignment of activity to industries by reporting unit is not going to be possible. Thus a combination of disintegration and cross-country knowledge flows make the attribution of advertising to industries a matter of caution which it is hoped that future work will address. 20

21 2.3.2 Market Research Our previous work simply used the turnover of the market research industry (SIC74.13, 2.3bn, then doubled to reflect own account). The main source of data here to estimate market research by industry is the Use Table which provides the intermediate consumption in market research and management consultancy (product group 111) split by sector (123 industry detail). In addition, we add in the product group 111 in the Gross Fixed Capital Formation (GFCF) by industry IO table. The product group 111 includes market research and management consultancy data. We are able to compute a market research share by using data on market research and management consultancy value added from the ABI. We then apply this share to the intermediate consumption and GFCF by industry data. Afterwards we gather the data in order to get our six categories. Finally we double the figures in order to consider the own account market research. Figure shows the market research expenditure for all the sectors. It is possible to see how the expenditure on market research has grown up in the resent years. In 2004 the total private sector spending on market research is 4.95bn. The financial and business sector is responsible for more than half of this total. 21

22 Figure Market Research by sector Tot FinBsSvc RetHtTrn Mfr Cons year Tot Mfr Cons FinBsSvc AgMin Util RetHtTrn Comparing these results with GHW we find similar estimates: in GHW the total expenditure on market research for 2004 is 4.5bn, in the current work it is 4.95bn FIRM-SPECIFIC HUMAN CAPITAL Most of the available data on training provides an indicator of whether the respondent received training. Other surveys tend to focus on skill shortages experienced by employers and hard-to-fill vacancies as perceived by employers. Collecting data on cost of employer-provided training is more complicated since the cost of employer-provided training not only encompasses the cost of providing training but also the opportunity costs of paid employee time whilst undergoing training. We estimate the training expenditure by using the data from the National Employer Skills Survey 2004 (NESS2004). The NESS obtains training expenditure in two stages: first it surveys a large number of firms (74,500 employers) to see if they are training or not and second they conduct a separate follow-up inquiry on the firms 22

23 who say they had funded or arranged training in the previous 12 months (who represents the 65% of the interviewed employers in the first step). This second survey is more detailed and estimates employer expenditure on training. The sample of employers for this second study is representative of the profile of training employers from the main survey by size, sector and the type of training the establishments provides (off-the-job training only, on-the job-training only or both type of training). The cost of training involves labour costs of those receiving and delivering or organising training. In particular, on-the-job training cost includes trainee and trainers labour cost. Off-the-job training includes trainee labour cost, fees to external providers, training management, travel and subsistence etc The NESS survey provides the training expenditure figures by sector skills council (SSC). SSCs are the employer-led organizations charged with leading the skills and productivity drive in sectors recognised by employer. One limitation is that this split is provided just for the year We build a training expenditure time series at an industry level by backcasting using two main sources: - the training expenditure in the 27 sectors classification for The EU KLEMS wage bill time series The idea is to construct a constant incidence time series which will show changes in training according to changes in industry composition. We proceed by firstly computing the sector-specific incidence for 2004 and secondly by applying the incidence to the wage bill time series available for each sector. A constant incidence time series could misrepresent the effective training expenditure trend. In fact one could expect the training incidence to increase over time for two reasons: Firstly, the fraction of professionals, who receive a lot of training, has gone up since the 1960s. Secondly, the size of the firms has risen and large firms are those who train more. As a result, we follow Nakamura s assumption that the incidence has been growing by 2% since 1956 (so it doubles in 50 years). Thus we apply to our time series a discount rate of 2%. 23

24 Finally, since the NESS survey covers just England, we apply the ratio of the employees receiving training in UK and in England to the training expenditure time series obtained above. Figure shows the training expenditure by sector. The manufacturing training expenditure does not present significant growth. Figure Training expenditure by sector Tot ReHtTrn FinBsSvc Mfr Cons year Tot Mfr Cons FinBsSvc AgMin Util RetHtTrn Comparing these results with GHW we find very similar estimates: in GHW the total expenditure on training for 2004 is 28.8bn, in the current work it is 28.15bn ORGANIZATIONAL STRUCTURE Organizational structure also presents us with challenge of accounting for purchased and own-account investment. The own-account component is assumed to be the value of the executive time spent on improving the effectiveness of business 24

25 organizations. The purchased component is represented by the management consultant fees. The expenditure on purchased management is estimated as the revenues of the management consulting industry. The main data source is the survey set up by the UK Management Consulting Association (MCA) of 64 firms in the UK consulting industries. They estimate their members are 70% of the industry and put their members fee income, in 2005, at 7.66bn, giving an estimated industry turnover of 11.9 billion. The MCA survey includes management consulting services (for example businessprocess re-engineering, strategy, change management) outsourcing-related consulting and IT-related consulting. It might be that some expenditure double count with investment or are destined to activities too short-lived to be asset-building. Given the difficulty in understanding how much of this expenditure is investment, for the moment we left them as they are. The MCA report provides an industry breakdown of 2005 fee income. After converting the available split in to our common classification, we construct the sector-specific expenditure time series by adopting the same procedure we followed for the advertising and market research expenditure time series. Assuming a constant incidence of the sector-specific expenditure on the private sector total expenditure, we compute the sector shares for the year 2005 and applied them to the estimated time series of the private sector expenditure on management consulting. In order to consider the consulting services purchased from abroad we add the management consultancy imports. As the MCA figures include the consulting services purchased by UK firms only we do not need to subtract the exports. Figure a shows the expenditure by sectors on management in the period Comparing these results with GHW we find similar estimates: in GHW the total expenditure on purchased organizational structure for 2004 is 7bn, in the current work it is 6.53bn (before the trade adjustment it is valued 4.3bn). 25

26 The share of bought-in management consulting taken by manufacturing, computed on this basis, is 41%. This differs from the share implied by data in ONS Input- Output tables. The basis of the difference remains to be investigated. Fig a Purchased Organizational structure by sector Tot Mfr S5&S6 Cons year Tot Mfr Cons S6=FinBsSvc AgMin Util S5=RetHtTrn The expenditure on own-account organizational structure is estimated as 20% of the managers earnings. We are following the CHS assumption that 20% of time is spent on organization building activities. Our source of data is the ASHE (Annual Survey of Hours and Earnings). In the appendix we present the list of the SOC codes of the managerial occupations we have considered for estimating this spending. In 2000 the 26

27 SOC 1990 was revised and updated and published as SOC For this, in the years we refer to the occupations codes of SOC For the period we refer to the occupations codes of SOC In our estimation we exclude the managers consulting industry SIC as is it is already included in the purchased organizational structure figures. The earnings are computed as the average gross annual earnings, thus they include incentives, additional premium payments and overtime pay. Fig b shows the expenditure on own-account organizational structure by sector. In almost all sectors the expenditure presents an increasing trend. Fig b Own account organizational structure by sector Tot S6 S5 S year Tot S2=Mfr Cons S6=FinBsSvc AgFish Util S5=RetHtTrn Comparing these results with GHW we find different estimates: in GHW the total expenditure on purchased organizational structure for 2004 is 15.3bn, in the current work it is 17.62bn. This is due mainly to a revision in the choice of the managers occupations codes. 27

28 3. FURTHER ANALYSIS OF THE MANUFACTURG SECTOR In this section we try to answer the following question: how much does the manufacturing spend on intangible assets relative to tangible assets? How much of the total investment in intangible assets is undertaken by manufacturing? We have to be careful here on matters of definition. We use the EUKLEMS data on investment and measure tangible investment as the sum of investment in ICT, excluding software, transport equipment, buildings, machinery and other. 4 In addition, our data above are for expenditures on various intangible categories. To go from expenditure to investment we follow CHS and assume that all such expenditures are investment with the exception of advertising (0.6) and bought in organisational capital (0.8). We also apply a factor of 0.5 to our design data, see above. Table 5 shows some key variables. Column 1 shows the ratio of manufacturing intangible investment to manufacturing tangible investment. Spending by manufacturing on intangible investment is now more than twice that on tangible investment. The second column of shows the ratio of manufacturing intangible investment to total intangible investment and shows that is has been falling. However as columns 3 and 4 show the share of manufacturing value added and employment in the total are also falling. Figure 3.1 shows the resulting shares of tangible and intangible investment in manufacturing value added time (manufacturing value added is the raw data from EUKLEMS and not adjust for the capitlisation of intangibles). The lowest line shows that tangible investment in manufacturing divided by value added in manufacturing has been falling. 5 The middle line shows the ratio of intangible 4 As a matter of fact, these data were close to the ONS manufacturing investment data published in the Monthly Digest. 5 Note that both these series are nominal. An alternative approach might be to compare the real investment to real value added or, say real investment. That would of course answer a different question. The nominal/nominal comparison asks about the share of all investment. The real/ real comparison asks what share of tangible investment the business sector would have to spend on investment to buy those investment capital goods that it bought in the particular year, at the prices they faced for those goods in some base period. Since the prices for some intangibles have been falling, driven a lot by software, this has of course been rising. Now, if investment prices have been falling then it is of course true that the same ratio of investment would buy more or better goods. 28

29 investment in manufacturing to value added in manufacturing has been rising. The top line shows the total, which is slightly falling. Fig 3.1: Ratios of manufacturing tangible investment, manufacturing intangible investment and their sum, to manufacturing value added year mfr intang inv/mfr va mfr total inv/mfr va mfr tang inv/mfr va Note: manufacturing tangible investment is manufacturing ONS measured investment less our measure of manufacturing software. 4. CONCLUSION We attempted to measure intangible investment by industry for the UK, using some new methods relative to previous work particularly in advertising and design. We present data on expenditure on intangible assets for the UK market sector for six industries over the last decade or so (data availability allowing). The sectors are (1) Agriculture, Fishing and Mining; (2) Manufacturing; (3) Electricity, Gas and Water; (4) Construction; (5) Wholesale and Retail, Hotels and Restaurants, Transport and Communications; (6) Financial Intermediation and Business Services. We use new methods relative to previous work particularly in advertising and design. Our main findings are as follows. First, overall intangible expenditure was, in 2004, around 154bn (investment was 92bn). Second, to give an illustration of the manufacturing 29

30 results, in 2004 manufacturing intangible investment was 40bn and tangible 12bn, a ratio of 2.5 to 1 up from 1.2 to 1 in Third, manufacturing accounted in 2004 for 12% of total tangible investment and 31% of total intangible investment. 30

31 References Sumiye Okubo, Carol A. Robbins, Carol E. Moylan, Brian K. Sliker, Laura I. Schultz, and Lisa S. Mataloni (2006), "BEA s 2006 Research and Development Satellite Account: Preliminary Estimates for ; Effect on GDP and Other Measures" available at < National Accounts, Concepts, Sources and Methods (2006), available at ( hods.pdf) 31

32 Table 1: Sources & method VESTMENT ITEM Computerized information SOURCES & METHODS DUSTRY BREAKDOWN AVAILABILITY PERIOD AVAILABLE Computer software ONS estimates 123 industries breakdown Own Account & Purchased: Computerized databases Included in software estimates Included in software estimates As above Innovative property Scientific R&D ONS Business Performed R&D (BERD) Main product breakdown Mineral exploration National Accounts, ONS series Series for coal, UKCSC oil and other mineral Copyright and license costs National Accounts, ONS series Series for publishing, public and broadcasting/recording consumption Other product development, design and research New product development costs in the financial industry New architectural and engineering designs Estimated as 20% of Financial Services industry's intermediate purchases (ONS data), discounting advertising, software, consulting and design purchases. Purchased are estimated for the IO tables.own-account are estimated using designers' earnings from ASHE data All in financial intermediation services Detailed industry breakdown available in ASHE data R&D in social science and humanities Economic competencies Brand equity Estimated as twice industry revenues of social science and humanities R&D industry no breakdown (73.2) from services inquiry (now ABI) Advertising expenditure Estimated from the IO tables. 123 Industry breakdown Market research Estimated from the IO tables. We then double the figures to consider the ownaccount 123 industries breakdown Firm-specific human capital Organizational structure NESS05, direct & indirect costs of employer provided training, adjusted to all UK NESS05 specific 27 industry split 2005 Purchased Own account Data on revenues of management consulting industry from Management Consulting Assocation. Estimated as proportion of managers' earnings using ASHE Specific industry breakdown provided by the management and consulting association Detailed industry breakdowon

33 Tab.2 Sectoral classifications for key intangible items SIC codes grouped as supply use tables coding (123 industries) Training R&D Organizational Structure (purchased) 1 1t5 1t5 1t t14 10t t16 15t16 15t to to t19 17t to t21 20t

34 t (except 24.4) to to t t to

35 t (except ) t t41 40t41 40t t52 50t55 50t t64 60t63 60t t67 65t t (pt) 70t74 70t (pt) is above to to

36 Tab.3 Common industry disaggregations available PROPOSED SECTOR CATEGORIES SIC CODES (2 DIGIT) NACE SECTIONS A17 1 1t14 A+B+C DUSTRY DESCRIPTION Agriculture, Fishing, Mining 2 15t37 D Manufacturing 3 40t41 E Electricity, Gas & Water Supply 4 45 F Construction 5 50t64 G+H+I Wholesale & Retail Trade, Hotels & Restaurants, Transport & Communications 6 65t74 J+K Financial Intermediation, Business Services 36

37 Tab.4 Expenditure by sector on key intangible assets in 2004 Intangible Sector Expenditure bn Sector expenditure as a percentage of the total spending on this intangible Software 1 AgMin % 2 Mfr % 3 Util % 4 Cons % 5 RetHtTrn % 6 FinBsSvc % total Scientific R&D New architectural & engineering design Advertising Market Research Firm-specific Human Capital 1 AgMin % 2 Mfr % 3 Util % 4 Cons % 5 RetHtTrn % 6 FinBsSvc % total AgMin % 2 Mfr % 3 Util % 4 Cons % 5 RetHtTrn % 6 FinBsSvc % total AgMin % 2 Mfr % 3 Util % 4 Cons % 5 RetHtTrn % 6 FinBsSvc % total AgMin % 2 Mfr % 3 Util % 4 Cons % 5 RetHtTrn % 6 FinBsSvc % total AgMin % 2 Mfr % 3 Util % 4 Cons % 5 RetHtTrn % 37